Aurelia Molzer

I believe our OSS-Fuzz results on `buf_independent`, which I've seen as deprioritized for now, point to this same underlying issue:  ``` Result #0: Err(Format(FormatError { inner: CorruptFlateStream { err:...

Git bisect points to beb6dcb330993d9ac84322dd177ee54507d2e3e8 as the commit causing the fuzz failure which does not make a whole lot of sense to me. Or in any case, this caused some...

Thanks for the detailed performance graphs. Closing this as a special case of #1601 (and yes, that's how long a fundamental design problem with traits sadly persists).

Ah sorry, wait, this one has a draft and the other one does not. I'll switch it around.

Different idea: Move away from an `enum` to an opaque `struct` and provide some `into_xxx(self) -> Result` (only if exact match), `as_xxx(&self) -> Option`, `convert_to_xxx(self) -> image::Result` (with conversions where...

Since you've done the std transform in many crates, one central question is how are we to maintain this? As in how does this affect the ability to accept patches...

If you want to read over `Vec` and `&'_ [u8]` only, I might suggest [`zune-image`](https://docs.rs/zune-image/0.4.15/zune_image/image/struct.Image.html#method.read). The implementation quality is solid and the author knows what they're doing. As coincidence would...

A note, instead of building an instance of in your framework's `Request` type just to satisfy the trait bound you might as well introduce your own type that contains just...

I tried drafting an interface once that would allow us to optimize certain cases such as rows/columns of contiuous pixels: see #866. But not currently in the library and nowhere...

@lazypassion In the case of `ImageBuffer`/`DynamicImage` the iterator is as optimized as iterating yourself. I have a feeling it might not be as fast as iterating over a pure slice...